ANALYSIS OF REALIZATION OF MINIMAL ALGORITHM IN CLUSTER ARCHITECTURE
The article is concerned about the theorem of minimal algorithm as applied to the problem of tuples sorting in the computer cluster. A more detailed proof of the theorem is provided. It eliminated the revealed discrepancies. Common TeraSort sorting algorithm is analyzed in the MapReduce system, which consists of two tasks. Upper estimates of the first task implementation time for each phase (Map, Shuffle, Reduce) in the Hadoop system were derived. It is shown that properties of minimal algorithm are violated under fixed probability of tuples filtering.
Keywords: minimal algorithm, sorting, MapReduce, Hadoop